Soil Saturated Hydraulic Conductivity Assessment From Expert Evaluation

7/29/2019 Soil Saturated Hydraulic Conductivity Assessment From Expert Evaluation

1/12

Soil saturated hydraulic conductivity assessment from expert evaluationof field characteristics using an ordered logistic regression model

Florencio Ingelmo a,b, M Jose Molina a, Jose Miguel de Paz b, Fernando Visconti a,b,*aCentro de Investigaciones sobre Desertificacion-CIDE (CSIC, UVEG, GV), Crta. Moncada-Naquera Km 4.5, 46113 Moncada, Valencia, Spainb Instituto Valenciano de Investigaciones Agrarias-IVIA, Centro para el Desarrollo de la Agricultura Sostenible-CDAS, Crta. Moncada-Naquera Km 4.5, 46113 Moncada,

Valencia, Spain

1. Introduction

Saturatedhydraulicconductivity (Ks),asameasureoftheabilityof

soil to transmit water, is essential in infiltration-related applications

such as irrigation and drainage management (Wu et al., 1999;

Radcliffe and Rasmussen, 2002) and for modelling the hydrology of

the landscape. This parameter is obviously related to the hazard of

pondingandto thepotentialofsoilsfor tiledrainage,whichcanaffectthe production of certain crops (McKeague et al., 1982).

Ring infiltrometers are often used for measuring the water

intake rate at the soil surface. The totalflow rate into the soil from a

single-ring infiltrometer is a combination of both vertical and

horizontal flow.Wuet al. (1997) found that the infiltration rate of a

single-ring infiltrometer was related to the one-dimensional (1-D)

infiltration rate for the same soil. For a relatively small ponded

head, the 1-D final infiltration rate of a field soil is approximately

equal to the field Ks, which is valuable information for computer

modelling and irrigation management.

Even with improved equipment, the Ks measurement is time

consuming, and thus, models are recommended. Several attempts

(Rawls et al., 1982, 1998; Tietje and Hennings, 1996; Dexter andRichard, 2009) have been made to estimate the Ks from readily

available analytical soil data such as particle size distribution, bulk

density and organic matter content by means of pedotransfer

functions or by physical modelling of the pore size distributions.

However, all these estimation methods exhibit large differences

betweenpredictions andmeasurements of Ks (Tietje andHennings,

1996; Landini et al., 2007), or the hydraulic conductivity close to

water saturation could not be estimated based only on the usually

available estimators (Weynants et al., 2009). Models based on soil

characteristics such as bulk density and pore size distribution give

better predictions as shown by Mbagwu (1995), who estimated Ks

Soil & Tillage Research 115116 (2011) 2738

A R T I C L E I N F O

Article history:Received 6 October 2010

Received in revised form 27 May 2011

Accepted 12 June 2011

Keywords:

Soil hydrology

Soil hydraulic conductivity

Ordered logistic regression model

Correspondence analysis

Hydropedology

A B S T R A C T

The knowledge of the soil saturated hydraulic conductivity (Ks) is essential for irrigation managementpurposes and for hydrological modelling. Several attempts have been done to estimate Ks in base of a

number of soilparameters.However, a reliable enoughmodel for qualitativeKs estimation based on the

expert assessment of field characteristics had not been developed up to date. Five field characteristics,

namelymacroporosity (M), stoniness (S), texture(T), compaction(C) andsealing(L), in addition to tillage

(G)were carefully assessedaccordingto three classeseach, in 202sites in anagricultural irrigatedareain

EasternMediterranean Spain. After the evaluationof field characteristics, a single ring infiltrometer was

used to determine the Ks value as the solution of the infiltration equation when the steady state was

reached. The distribution of the Kswas assessed andfive classes with 10-fold separations in class limits

were defined accordingly. The relationships among site characteristics and Ks were analyzed through a

correspondence analysis (CA). Next, an ordered logistic regression model (OLRM) for the prediction of

the Ks class wasdeveloped. TheCA revealed that, though tightly related, the set of six sitecharacteristics

should not be simplified into a smaller set, because each characteristic explains a significantly different

aspect ofKs. Consequently, theOLRMwasbased on thesix characteristics,whichpresented the following

order of importance: L>M> G> T > C > S. According to the cross-validation of the OLRM the hit

probability for the prediction of theKs classattainedan average valueof 50%, which increased to 63% forthe highest class ofKs. Moreover, wrong estimation of the Ks class exceeded the 1 range only in 3% of

sites. Therefore, a reliable enough assessment of Ks can be based on the expert assessment of field

characteristics in combination with an OLRM.

2011 Elsevier B.V. All rights reserved.

* Corresponding author at: Centro de Investigaciones sobre Desertificacion-CIDE

(CSIC, UVEG, GV), Crta. Moncada-Naquera Km 4.5, 46113 Moncada, Valencia, Spain.

Tel.: +34 963 424 000; fax: +34 963 424 001.

E-mail address: [email protected] (F. Visconti).

Contents lists available at ScienceDirect

Soil & Tillage Research

journal homepage : www.elsev ier .co m/loc ate /s t i l l

0167-1987/$ see front matter 2011 Elsevier B.V. All rights reserved.

doi:10.1016/j.still.2011.06.004
http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004mailto:[email protected]://www.sciencedirect.com/science/journal/01671987http://dx.doi.org/10.1016/j.still.2011.06.004http://dx.doi.org/10.1016/j.still.2011.06.004http://www.sciencedirect.com/science/journal/01671987mailto:[email protected]://dx.doi.org/10.1016/j.still.2011.06.004


2/12

from bulk density, macroporosity, mesoporosity and microporosi-

ty. Since thesemodels are generallyhighdatademanding and need

cumbersome laboratory determinations, the applicability for

farmers in irrigation management is reduced. To avoid this,

qualitative models based on the expert assessment of morpholog-

ical characteristics of soil could be an alternative approach to

model the Ks. This idea of qualitatively describing water flow

through soils has been credited to the Soil Conservation Survey

(Norton, 1939). Since then, several models for the qualitative

classification of soil ease to permit water flow have been

developed. Mason et al. (1957) developed such a model based

on the expert assessment of 14 soil morphologic characteristics in

order to classify 900 soils in an ordinal scale of seven permeability

classes, defined as the ease with which pores of a saturated soil

permit water movement. They attained a hit probability of 30%,

and suggested that 95% probability of making a correct prediction

could be achieved by using only three to five permeability classes.

McKeague et al. (1982) developed guidelines for estimating the

class of saturated hydraulic conductivity of soil horizons from

observations of soil morphology in 78 soil horizons ranging in

texture from sandy to clayey. The major factors contributing to

high Ks values were abundant biopores, textures coarser than

loamy fine sand, and strong, fine to medium blocky structure. The

lowest values were associated with clayey horizons that had beencompressed or puddled by cultivation. The guidelines presented,

though incomplete and subjective to some degree, improved the

estimates of Ks in limited testing by pedologists. The results also

indicated that it was not feasible to assign a unique Ks estimate to

near-surface horizons of cultivated soils of a particular series.

Tillage practices and current land use have a major effect on soil

structure, porosity and density, and hence on Ks.

Saturated hydraulic conductivity can also be related to soil

morphological criteria based on the expert assessment and the

classes of the Factual Key (McKenzie et al., 2000). Lin et al. (2006)

presented a vision that advocates hydropedology as an advanta-

geous integration of pedology and hydrology for studying the

intimate relationships between soil, landscape, and hydrology.

Landscape water flux is suggested as a unifying precept forhydropedology, through which pedologic and hydrologic exper-

tise can be better integrated. The discretization of continuous

field measurements such as the Ks, is usually of high practical

value to perform this integration. The indication of a class of

Ks is, on the one hand, more informative, and on the other hand,

more stable in space and time than the indication of an X%

confidence interval derived from an ordinary least squares

regression model.

Given a saturated hydraulic conductivity expressed in an

ordinal scale, the datum to predict is not longer the actual value of

Ks, but the probability of an observation to belong to a certain class

of Ks. This can be adequately performed using logistic regression

models (LRM). Logistic regression modelling has been previously

used

in

soil

research

to

assess

water

erosion

from

expertevaluation of site characteristics (Sonneveld and Albersen,

1999). The development of a LRM appears as an adequate

methodology for predicting an ordinal variable from other ordinal

variables, which to our knowledge has not been carried out up to

date for the Ks assessment.

The objective of the present study was to develop a

methodology for the estimation of the class of soil saturated

hydraulic conductivity based on several field characteristics such

as tillage, macroporosity, stoniness, texture, compaction and

sealing. This main objective was split into two partial objectives:

(i) the development of a methodology for the expert evaluation of

the soil characteristics, and (ii) the development of an ordered

logistic regression model for the Ks prediction on basis the six field

characteristics.

2. Materials and methods

2.1. Study area

The study area (Fig. 1) has 12,400 ha, of which approximately

6300 ha are agricultural irrigated lands. Citrus is the main crop

with 53% of the irrigated area, followed by vegetables (mainly

melon and watermelon) with 14%. Rice crop in lands with shallow

watertables accounts for no more than 3% of the area. Citrus

orchards and some vegetables are generally drip irrigated. Drip

irrigation is used on 65% of the total irrigated area. The climate can

be considered as semiarid following the UNESCO classification (De

Paw et al., 2000), with annual rainfall of 500 mm and reference

evapotranspiration (ET0) of 1000 mm.

In regard on landscape features, three main areas, associated

with soil types, can be distinguished: (i) the colluvial and glacis

area, with soil materials moved, accumulated, removed or even

replaced, especially when calcareous duricrusts can limit the

effective soil depth for citrus crops; typical soils there are,

respectively, aric Anthrosols and petric Calcisols; (ii) the flood-

plains and alluvial area, with more fertile finer-textured soils such

as Luvisols, which are typically cultivated for citrus; (iii) the third

area located near the coast, characterized by lacustrine fine-

grained deposits from infilling of lakes and early used for rice andhorticultural crops. In this last area, a watertable is present

seasonally, and subsurface horizons of lacustrine soils often show

visible greyish colours and prominent reddish mottles when

oxidising conditions occur periodically by tile drainage.

2.2. Soil survey

The surveyed plots were selected according to a combined

systematic and random point selection in agreement with De Paz

et al. (2011), in order to have all soil types and crops represented

according to their predominance in the area. The study was carried

out during the irrigation season, and in days when the soil water

content was close to field capacity, i.e., between 1 and 3 days after

irrigation. In each of the 101 plots (Fig. 2), two separated pointswere selected for a total of 202 survey sites. In each site several

10 cm 10 cm areas with no vegetation were delimited for the

expert assessment of field characteristics, thereafter infiltration

was measured in one of the areas.

2.3. Saturated hydraulic conductivity determination

The soil water infiltration rate was measured using a single

head ring infiltrometer according to the method by Wu et al.

(1999). The single-ring infiltrometer consisted of an infiltration

ring 12 cm indiameterand6.5 cm inheight with a calibratedwater

supply column that maintains a constant water pressure head of

1 cm on the soil inside the ring (Fig. 3). The ring insertion depth in

the

soil

was

5

cm.

The

cumulative

infiltration

at

different

timeswas measured by annotating the height of water in the water

supply column. The time of measurement was sufficiently long to

achieve a steady-state infiltration rate, which was usually 20 min.

Calculations of saturated hydraulic conductivity were obtained

from Eqs. (1)(3) (Wu et al., 1999),

Ks A

af (1)

f %H 1=a

G 1 (2)

G d r

2

(3)

F. Ingelmo et al./ Soil & Tillage Research 115116 (2011) 273828


3/12

where A is the slope of the cumulative infiltration (cm) vs. time

(min) curve, a is a constant with value close to 1, H is the constant

ponded head of the infiltrometer (1 cm), a values were assumed tobe 0.36 cm for coarse textured soils, 0.12 cm for loamy soils, and

0.04 cm for clayey soils (Elrick and Reynolds, 1992), d is the depth

of insertion of the ring (5 cm), and r is the radius of the ring (6 cm).

2.4. Assessment of the field characteristics

The field characteristics credited to be most related to the

hydraulic behavior of soils according to Hillel (1998), Porta et al.

(1994) and Radcliffe and Rasmussen (2002), include tillage (G),

macroporosity (M), texture (T), compaction (C), sealing (L), and

stoniness (S). These characteristics were assessed as qualitative

parameters by visual (G, M, L, S) and feeling to touch (T) methods(Porta et al., 1994; Milford et al., 2001; Ball and Douglas, 2003;

Mueller et al., 2009), and by the resistance to the insertion of the

ring of the infiltrometer (C). The results of the visually evaluated

characteristics macroporosity and stoniness could be validated by

image processing. Although this validation was not formally

carried out, the expert classification of macroporosity and

stoniness was found to be in accordance to the results of the

image processing performed on the zenithal photographs of the

soil surface taken in some fields.

2.4.1. Tillage

Management and tillage practices have significant influence on

different hydraulic properties, because under agricultural land use,

the

properties

of

the

macropore

system

strongly

depend

on

themamong others factors (Wahl et al., 2003). Particularly, the saturated

hydraulic conductivity of soils is dominated by the micromorphol-

ogyof soil pores rather thanby the total porosity. Micromorphology

of pores in topsoils is subjected to continuous disturbance by

frequent tillage, while subsoils tend to be compacted without

serious changes of micromorphology of soil pores (Nakano and

Miyazaki, 2005). Two classes of tillage were defined:

Class 1 or tilled soils. The soil surface clearly shows tillage

practices (Fig.4 left).Thisalways occurs in vegetable gardens and

in some citrus orchards with tillage for weeds control. Class 0 or non-tilled soils. This kind of soil management occurs

for most soils under citrus, either drip or surface irrigated (Fig. 4

right).

Fig. 1. Location of the study area.

Fig.

2.

Distribution

of

the

101

selected

plots

in

the

study

area.

F. Ingelmo et al./ Soil & Tillage Research 115116 (2011) 2738 29


4/12

2.4.2. Texture

Texture was evaluated in the field according to the feeling to

touch, the stickiness and the internal cohesion of the soil particles

according to the method described by Milford et al. (2001), e.g., a

soil sample of about 25 g was held on the palm of the hand, then it

was moldered to form a ball or bolus, adding soil or water until the

bolus failed to stick to the fingers. The behavior of the bolus and of

the ribbon produced by pressing out between thumb and

forefinger and the feel of the material between both fingers were

used to classify texture. Three textural classes were used, each one

representing a general increment from more to less ease for soil

tillage:

coarse

(1),

medium

(2)

and

fine

(3).

In

Table

1

thecorrespondences between the textural classes obtained with the

Milford et al. (2001) method and the USDA classes are shown.

2.4.3. Compaction

Soil compaction affects hydraulic properties, and thus can lead

to soil degradation and other adverse effects on environmental

quality (Zhang et al., 2006). Soil compaction changes the ability of

soil to hold water, decreases infiltration rate and saturated

hydraulic conductivity, and increases penetration resistance

(Shafiq et al., 1994). Given that the penetration resistance

decreases with water content and increases with bulk density, a

qualitative field evaluation of soil compaction was assessed during

the insertion of the ring of the infiltrometer in a fresh vertical

exposure

when

the

soil

was

near

field

capacity.

The

strength

of

the

soil wasjudged from the time required to insert the ring just 5 cm

into the soil when the compressive force of a nylon hammer

applied at a height of 50 cm strikes on the coverlid of the ring,

which is equal in diameter. This estimate of compaction could be

given from measurement of the bulk density, but could not be

substituted by a penetrometer resistance measurement. The use of

the ring was preferred over a penetrometer because the former

gives averaged information of the area subsequently used for the

infiltration assay, while the latter gives just point information,

which is very variable. Depending on the time needed to insert the

cylinder the compaction class was estimated as follows:

Class 1 or low compaction: the time was less than 1 min. Class 2 or medium compaction: the time was within 13 min. Class 3 or high compaction: the time was longer than 3 min.

Fig. 3. Scheme and field photograph of the constant head ring infiltrometer used to carry out the infiltration measurements.

Fig. 4. Field examples of tillage in the study area.

Table 1

Correspondence between the textural classes obtained with the Milford et al.

(2001) method and the USDA texture classes.

Class with

Milford et al.

(2001) method

Equivalent USDA texture class

1 Sand, loamy sand, clayey sand, sandy loam

2 Loam, silty loam, sandy clay loam, silty clay loam

3 Sandy clay, silty clay, clay loam, clay



5/12

2.4.4. Macroporosity

Macroporosity is defined as the abundance of voids with

diameters between 1 and 5 mm. It includes the modification of the

soil surface morphology, in particular structure, either by soil

fauna, mainly microarthropods and earthworms (Weiler and Naef,

2003), plant roots (Holden, 2005) or by permanent cracks between

peds that usually do not close on wetting. Despite their low

abundance, macropores can have a significant impact on infiltra-

tion and runoff (Jarvis, 2007), and play an important role in the

rapid transmission of water, known as preferential or bypass flow,

in

both

the

topsoil

and

the

subsoil

(Franklin

et

al.,

2007). It

mayresult in the lowering of the bulk density and subsequent

increasing of the water infiltration rate. Therefore, the visible

abundance of both biota-made macropores (circular voids, Fig. 5

top row) and cracks (planar voids, Fig. 5 bottom row) is jointly

described (Dexter and Richard, 2009). Three classes of macro-

porosity in the site areas of 10 cm 10 cm were evaluated

according to Fig. 6, which is based on several results from the

literature (Lachnicht et al., 1997; Buczko et al., 2006; Holden and

Gell, 2009):

Class 1 or low macroporosity: 03 voids/dm2. Class 2 or medium macroporosity: 36 voids/dm2. Class 3 or high macroporosity: more than 6 voids/dm2.

2.4.5. Sealing

Unstable materials high in silt and fine sand, or fine textured

well-structured soils may slake, swell and disperse. Under such

surface conditions, porosity and water infiltration decrease (Laj

et al., 2001). The sealing variable is a simple field assessment of

the effects of aggregate instability and dispersibility induced by

texture and structure conditions of soil as well as by irrigation

water quality resulting in the accumulation of fine silt and fine

sand or dispersed clay and salts on the soil surface forming seals

and crusts of mechanical origin (Fig. 7 left). Under shaded areas

of

drip-fertirrigated

citrus, this

kind

of

seals

and crusts

enhancealgae, mosses and lichens colonization, and contribute to form

biological crusts that usually are water repellent (Fig. 7, center

and right). Both types of seals have been jointly evaluated. Three

sealing classes were defined according to seal thickness and the

percentage of covered soil surface (Fig. 8) in the site areas of

10 cm 10 cm:

Class 1 or low sealing: the seal is 5 mm thick and covers more

than the 60% of the soil surface.

Fig.

6.

Model

sheets

to

evaluate

the

soil

macroporosity:

the

black

points

represent

the

voids.

Fig. 5. Field examples of macroporosity evaluation in the study area.



6/12

2.4.6. Stoniness

Three classes of stoniness were estimated from the observation

of the percentage of the site areas of 10 cm 10 cm covered by

rock fragments according to Fig. 8 (Daniels et al., 1968; Poesen

et al., 1990):

Class 1 or low stoniness: 025% Class 2 or medium stoniness: 2550%

Class 3 or high stoniness: 5075%

Field examples of the three classes of stoniness are shown in

Fig. 9.

2.5. Correspondence analysis (CA)

The field characteristics are related to each other as has been

stated when describing their estimation in the previous section,

e.g., tillage, macropororosity and compaction. Therefore, a

correspondence analysis (CA) was applied to find out to what

extent the characteristics were associated and if the likely

information redundancies could be reduced.

An exploratory study of the multiple dimensional contingencytable of the n = 6 independent variables and Ks was carried out by

means ofCA. Particularly, thematrixused for such analysiswasnot

the contingency table itself but the matrix of scores of s sites n

Fig. 7. Field examples of sealing evaluation in the study area.

Fig.

9.

Field

examples

of

stoniness

evaluation

in

the

study

area.

Fig. 8. Model sheets to evaluate the surface cover of sealing and stone contents.



7/12

field characteristics. Similarly to principal components analysis

(PCA) the eigenvectors of the matrix were obtained in order, from

the one that accounted for the highest variance, or inertia in CA

terms, to the lowest. Then, the coordinates of the different classes

of the field characteristics and those of the Ks classes were

calculated, whichallowed toproject them ona Euclidean space and

interpret the relationships among them.

The CA was carried out using the statistical package R (The R

Development Core Team, 1999) and graphics drawn with a

Microsoft Excel spreadsheet.

2.6. The ordered logistic regression model (OLRM)

There are several approaches to logistic modelling. However,

when the outcome represents an underlying continuous scale

subdivided in several categories, as is the case with Ks, the most

adequate modelling framework is a cumulative approach (Full-

erton, 2009). The generalized ordered logistic model for the

cumulative approach, when modelling a variable z split in m + 1

categories from n independent variables, is given by the following

equations (Eqs. (4) and (5)).

y abx (4)

yi LogPrz!vi

Prz


8/12

The associations among the different classes of the explanatoryvariables were studied through correspondence analysis (CA)

including also the dependent variable Ks class. The first two factors

were the only having an associated eigenvalue higher than 0.5.

Two coordinates for each class of the site characteristics were

calculated to give a two dimensionalmap of variables (Fig. 12). The

first factor explained mainly the variability associated with tillage

and texture, and to a lesser extent to low and medium compaction

and medium and high macroporosity. Based on the cluster formed

by G1, T1 and C1, the tilled soils are often light textured and low

compacted and, given the proximity of S1 and L1, also usually havelow stone content and low sealing. The soil surface of low

compacted soils has usually continuous and clearly visible

connecting pores and cracks that do not close on wetting. The

structure presents strong, stable, cemented aggregates and

nodules and gravels are scarce. Given the cluster formed by G0,

T2, T3, C2, C3, S2 and S3, the non-tilled soils are predominantly

medium to heavy textured, medium to high compacted and

medium to high stony. The structure is moderate and blocky,

polyhedral peds are common. If coarser, the inter-particle voids are

filled with fine minerals. There are few visible (hand lens) pores

that conduct water when wet. The pores and channels which

remain open when wet are visible. Rock fragments are commonly

present.

On the other hand, the second factor explained the variabilityassociated to stoniness, low and medium macroporosity and

medium and high sealing. The variability of macroporosity, sealing

and stoniness is interestingly associated with the second factor.

Ks / mm h-1

log Ks

Fig. 10. Distribution of the saturated hydraulic conductivity in the study area.

0

20

40

60

80

100

< 0,3 [0,3, 3,6[ [3,6, 35 [ [35, 350[ 350observed class of Ks / mm h -1

Numberofobservations

wrong

right

Fig. 11. Number of wrong and right predictions for each class of observed Ksaccording to the cross-validation.

Table 2

Number of observations for each class of the site characteristics.

Variable Symbol Class Number of observations

Code Name (range if appropriate)

Tillage G 0 Non-tilled 159

1 Tilled 43

Texture T 1 Light 26

2 Medium 83

3 Heavy 93

Compaction P 1 Low 48

2 Medium 104

3 High 50

Macroporosity M 1 Low 57

2 Medium 93

3 High 52

Stoniness S 1 Low 102

2 Medium 25

3 High 75

Sealing L 1 Low 152

2 Medium 35

3 High 15

Sat. Hyd. Cond. K 1 Extremely low (Ks


9/12

Low macroporosity is associated with medium to high sealing,

medium macroporosity with low to medium sealing and high

macroporosity with low sealing. The association between macro-

porosity and stoniness, although weaker than the preceding one, is

also present. High and medium stoniness is associated with high

and medium macroporosity, whereas low stoniness is associated

with low macroporosity.

Stone fragment in soils modifies the pore space in the contact

with the surrounding fine-earth fraction (Sauer and Logsdon,

2002). This effect influences water flow by affecting hydraulic

continuity near fragment surfaces. These relatively subtle mor-

phological factors may have disproportionate impact on water

flow under near-saturation conditions in these soils, and variabili-ty in space and time in water infiltration rates could be mainly

attributed to the influence of the stone fragment content (Verbist

et al., 2009). Ravina and Magier (1984) showed that the increase in

coarse fragment content increased the resistance to compaction

and, consequently, preserved lower soil bulk density. It was

concluded that coarse fragments contributed to improved physical

conditions by acting as a skeleton, which resists soil compaction.

As suggestedby Ingelmo et al. (1994), larger volumes of largepores

in soils with coarse fragments, can lead to an increase of the

effective porosity of the fine earth fraction of the soil and to a

highest hydraulic conductivity between fragments.

Given the proximity among L3, M1, C3 and T3, high sealing is

associated with low macroporosity, high compaction and heavy

texture.

Nevertheless,

this

association

is

weaker

than

the

previous.In high compacted soils structural peds are indistinct and barely

observable, or soil occurs as a coherent mass showing no evidence

of any distinct arrangement of soil particles. There is absence of

visible (hand lens) pores that could conduct water when wet.

Cracks and macropores between peds, when present, close on

wetting.

The saturated hydraulic conductivity variability is accounted

for by both the first and the second factors, however, each factor

accounts for the variability among different Ks classes. The first

factor, which is mainly related to tillage, texture and compaction,

accounts for the variability from the medium to thehighest level of

Ks (3,4 and5). Thehighest level of saturatedhydraulic conductivity

(5) is almost only found in tilled soils (8 out of 9 soils, see

supplementary

material

1).

Nevertheless,

since

not

all

tilled

soils

(G1) exhibit an extremely high Ks (K5), the simultaneous presence

of other field characteristics is decisive for the appearance of the

highest level of Ks. These are light texture (T1) and low compaction

(C1), which are typical of tilled soils but can also occur under no

tillage.

The closeness between the high level of Ks (K4) and the highest

level of macroporosity (M3) suggests this last as the most

important characteristic for the appearance of high infiltration

rates. The appearance of high macroporosity occurs in 52 sites and

these are usually non-tilled (69%), however, independently of the

tillage practices, the appearance of high macroporosity is almost

always accompanied by medium to high levels of Ks (96%).

On the other hand, the second factor accounts for the variabilityfrom the medium to the lowest level of Ks (1, 2 and 3). Given the

proximity between K1 and L3 and also M1, the lowest levels of Ksare mainly found in highly sealed soils with low macroporosity.

The sixfieldcharacteristics,whichhave been sought tobecome

explanatory variables for the class of Ks, are related to each other

and with Ks as it has been shown. However, the cumulative

percentage of inertia,i.e., the percent of variance explained by the

first two factors was only 19% (Fig. 13). It only reachedmore than

50% when considering seven factors, which is the number of

variables used for the CA (6 explanatory+ Ks). This suggests that

no variable could be suppressed a priori from the model

estimation without loss of significant information. In the same

way, no other set of variables derived from the six field

characteristicsbymeans of

a

mathematical transformation couldbe used instead of them without loss of significant information.

Although tightlyrelated,eachfieldcharacteristicseems toexplain

a significant different part of Ks variability and, therefore, should

be kept in the model.

3.3. Model estimation

The five classes of saturated hydraulic conductivity were

expressed by means of four dummy variables as indicated in

Table3. Then, four logit regressionequationswere calibratedwith

the following structure (Eq. (9)).

lnKs

1 Ks

ai bgiG btiT bciC bmiM bsiS bliL (9)

T1

T3

C2

C3

M1

M2M3

S1

S2S3

L1L2

K1

K2

K3

K4

K5

G0

G1

T2

C1

L3

40-

30-

20-

10-

0

10

20

-50 -40 -30 -20 -10 0 10 20 30

103

factor 1

103

facto

r2

Tillage (G)

Texture (T)

Compaction (C)Macroporosity (M)

Stoniness (S)

Sealing (L)

Sat. Hyd. Cond. (K)

Fig. 12. Projection of the classes of thefield characteristics on the plane given by thefirst and second factors extracted from the correspondence analysis. Successive Ks classes

are linked by broken lines.



10/12

The estimation of the regression coefficients is shown in

Table 4. As we expected from experience and the CA, tillage,

macroporosityandstoninesshave positive regressioncoefficients

in every equation while the contrary occurs with texture,

compaction and sealing. Accordingly, the higher classes of Ksare more likely in tilled than non-tilled soils, and in soils with

higher macroporosity and stoniness. On the other hand, lower

levels of Ks were observed with finer texture, and higher

compaction and sealing. The magnitude, and where appropriate,the significance of the regression coefficients follow different

trends for each characteristic.

The tillage variable is not significant for explaining the

variability between the lowest class of hydraulic conductivity,

which was found mainly under no-tillage, and the rest of classes.

Nevertheless, the tillage is significant to distinguish between the

lowest (1, 2, 3) and the highest levels (4, 5) of Ks. Moreover, its

magnitude increases for the distinction between the extremely

high and the rest of levels.

The texture is not significant to distinguish the extremely low

(1) from the rest (2, 3, 4, 5), but it is significant to distinguish

among these. The compaction is not significant to explain the

variability between the extremely high and the rest of classes of

hydraulic

conductivity,

but

so

is

to

explain

the

variability

amongthese. Moreover, its magnitude increases as Ks decreases. The

macroporosity is not significant to distinguish the extremely low

class of Ks from the others. Nevertheless, its magnitude increases

for the distinction among the higher classes of Ks. The stoniness is

only significant for the distinction between the extremely low and

low classes of Ks from the others. Finally, the sealing is not

significant for the distinction between the extremely high class of

Ks and the others. However so is to explain the variability among

the lower classes, particularly to distinguish between the

extremely low to medium (1, 2, 3) and high to extremely high

(4, 5). The differences among the regression coefficients across the

logit equations were found to be significant according to the test of

parallel lines. Given this result a simplified model with a common

set of regression coefficients (bg, bt, bc, bm, bs, bl), which would be

based on the so called proportional odds hypothesis (parallel

regression) should not be adequate. Such a model was, therefore,

discarded.

According to the magnitude of the regression coefficients the

six field characteristics can be arranged from the most to

the least important on determining the class of Ks: sealing >

macroporosity > tillage> texture > compaction > stoniness.

3.4. Model self-evaluation

The class of saturated hydraulic conductivity in each site was

assessed calculating the probability of each class of Ks (see

supplementary material 2). The expected class of Ks was

considered to be, therefore, the most likely one. According tothe comparison of predictions and observations, the class of Kswas

correctly calculated for 125 out of 202 sites, i.e., 62% of right

calculations. The highest percentage of right calculations was

attained for the medium class of Ks (71%), followedby the low class

(68%), next the extremely high class (56%), and then the high class

(51%). Conversely, the extremely low class of Ks was the only

predicted with percentage less than 50%, to be precise 40%.

In71 out of77 sites (92%)where the class of Kswasnot correctly

calculated, thepredicted Ks classwas only one classhigher or lower

than the observed class. In the remaining six sites the predicted Kswas two classes higher than the observed one (extremely low and

low Ks) or two classes lower (medium Ks). There were not

differences higher than two between observed and predicted

classes of Ks.

3.5. Cross-validation of the model

According to the cross-validation the class of hydraulic

conductivity was independently assessed in 200 sites (see

supplementary material 3). The class of Kswas correctly predicted

in 99 sites, i.e., 50% of right calculations. As shown in Fig. 11, the

highestpercentage of right calculations was attained in thehighest

and medium classes of Ks with 63 and 59% of right calculations,

respectively. These were the only classes with hit percentages

higher than 50%. The low, high, and lowest classes of Ks were

predicted with hit percentages of 45, 38 and 33%, respectively.

In 95 out of the 101 sites where the class of Ks had not been

correctly

predicted

(94%)

the

difference

between

the

observed

andthe predicted class of Ks was one in absolute value. In the

remaining six sites the difference was two in absolute value.

Differences in absolute value of three or larger did not appear

(Table 5).

0

20

40

60

80

100

1 3 5 7 9 11 13 15factor

percentofinertia

0.0

0.10.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

eigenvalue Cumulative %

of inertia

Eigenvalue

Fig. 13. Cumulative percent of inertia and eigenvalue of the factors obtained in the

correspondence analysis.

Table 3

Codification algorithm for the saturated hydraulic conductivity as four dummy

variables.

Dummy

variable

Condition

for Ksi 1=mm h1

Condition

for Ksi 0=mm h1

Ks1 Ks!0.3 Ks


11/12

Finally, in 29 out of 200 sites the probability of the most likely

class of Ks did not surpass 0.5. In order to have an individual hit

probability of at least 0.5, more than one class of Ks had to be

predicted for these 29 cases. Only two classes of Ks were enough,

and the percentage of right predictions attained 57% (113 out

of 200).

4. Conclusions

A

methodology

for

the

qualitative

classification

of

the

soilsaturated hydraulic conductivity based on the evaluation of field

characteristics has been developed. Based on the current

knowledge on soil hydraulic conductivity, five field characteristics

namely, macroporosity, sealing, texture, compaction and stoni-

ness, in addition to tillage were selected, and a methodology for

their field expert classification in three classes each was

established. Particularly, a new methodology for compaction

assessment was described. According to a correspondence analy-

sis, several remarkable associations among the six field character-

istics were revealed. However, the characteristics explained

different aspects of soil saturated hydraulic conductivity, i.e., each

class acts at different levels of the Ks, and simplification into a

smaller set of variables would not have been adequate. Therefore,

the six characteristics were subsequently used as a set of six

independent variables to develop an ordered logistic regression

model for the estimation of five classes of Ks with 10-fold

separations in class limits. The cross-validation of the OLR model

gave a hit probability of 50% with error estimations seldom outside

the 1 range. Therefore, a reliable enough assessment of Ks can be

based on the expert assessment of field characteristics plus the use of

an ordered logistic regression model.

Acknowledgments

This work hasbeendone in the frameworkof projects CGL2006-

13233-CO2-01 and CGL2006-13233-CO2-02. The authors ac-

knowledge the Ministerio de Educacion y Ciencia from Spain their

financial

support.

We would

like

to

thank

the

two

anonymousreviewers for their constructive comments.

Appendix A. Supplementary data

Supplementary data associated with this article canbe found, in

the online version, at doi:10.1016/j.still.2011.06.004.

References

Ball,B.C.,Douglas,J.T., 2003.A simpleprocedure forassessing soil structural, rootingand surface conditions. Soil Use Manage. 19, 5056.

Buczko, U., Bens, O., Huettl, R.E., 2006. Tillage effects on hydraulic properties andmacroporosity in silty and sandy soils. Soil Sci. Soc. Am. J. 70, 19982007.

Daniels, R.B., Gamble, E.E., Bartelli, L.J., Nelson, L.A., 1968. Application of the point-

count

method

to

problems

of

soil

morphology.

Soil

Sci.

106,

149152.

DePaw, E., Gobel,W.,Adam, H.,2000.Agrometeorological aspectsof agricultureandforestry in the arid zones. Agric. Forest Meteorol. 103, 4348.

De Paz, J.M., Visconti, F., Rubio,J.L., 2011. Spatial evaluation of soil salinity using theWET sensor in the irrigated area of the Segura river lowland. J. Plant Nutr. SoilSci. 174, 103112.

Dexter, A.R., Richard, G., 2009. The saturated hydraulic conductivity of soils with n-modal pore size distributions. Geoderma 154, 7685.

Elrick, D.E., Reynolds, D.W., 1992. Infiltration from constant-head well permea-meters and infiltrometers. In: Toop, G.C., Reynolds, W.D. (Eds.), Advances inMeasurements of Soil Physical Properties: Bringing Theory into Practice. Amer-ican Society of Agronomy, Madison, WI, pp. 124.

Franklin,

D.H.,

West,

L.T.,

Radcliffe,

D.E.,

Hendrix,

P.F.,

2007.

Characteristics

andgenesis of preferential flow paths in a Piedmont ultisol. Soil Sci. Soc. Am. J. 71,752758.

Fullerton, A.S., 2009. A conceptual framework for ordered logistic regressionmodels. Sociol. Methods Res. 38, 306347.

Hillel, D., 1998. Environmental Soil Physics. Academic Press, New York, pp. 173201.

Holden, J., 2005. Piping and woody plants in peatlands: cause or effect? WaterResour. Res. 41, W06009.

Holden,J., Gell, F.K., 2009. Morphological characterization of solute flow in a brownearth grassland soil with cranely larvae burrows (leatherjackests). Geoderma152, 181186.

Ingelmo, F., Cuadrado, S., Ibanez, A., Hernandez,J., 1994. Hydric properties of someSpanish soils in relation to their rock fragment content implications for runoffand vegetation. Catena 23, 7385.

Jarvis,N.J., 2007.A review ofnon-equilibriumwater flowand solute transport in soilmacropores: principles, controlling factors and consequences for water quality.Eur. J. Soil Sci. 58, 523546.

Lachnicht, S.L., Parmelee, R.W., McCartney, D., Allen, M., 1997. Characteristics of

macroporosity

in

a

reduced

tillage

agroecosystem

with

manipulated

earth-wormpopulations: implications for infiltrationandnutrient transport. Soil Biol.Biochem. 29, 493498.

Laj, H.R., Ahuja, R., Green, T.R., Benjamin, J.G., 2001. Residue cover and surface-sealing effects on infiltration: numerical simulations for field applications. SoilSci. Soc. Am. J. 65, 853861.

Landini, A., Martnez, D., Dias, H., Soza, E., Agnes, D., Sainato, C., 2007. Modelos deinfiltracion y funciones de pedotransferencia aplicados a suelos de diferentetextura. Ciencia del Suelo (Argentina) 25, 123131.

Lin,H., Bouma,J., Pachepsky,Y., Western, A., Thompson,J., van Genuchten,R., Vogel,H.J., Lilly, A., 2006. Hydropedology: synergistic integration of pedology andhydrology. Water Resour. Res. 42, W05301.

Mason, D.D., Lutz, J.F., Petersen, R.G., 1957. Hydraulic conductivity as related tocertain soil properties in a number of great soil groups sampling errorsinvolved. Soil Sci. Soc. Am. Proc. 21, 554561.

Mbagwu, J.S.C., 1995. Saturated hydraulic conductivity in relation to physicalproperties of soils in the Nsukka Plains, Southeastern Nigeria. Geoderma 68,5166.

McKeague,J.A., Wang, C., Topp, G.C., 1982. Estimating saturated hydraulic conduc-

tivity from soil morphology. Soil Sci. Soc. Am. J. 46, 12391244.McKenzie, N.J., Jacquier, D.W., Ashton, L.J., Cresswell, H.P., 2000. Estimation of Soil

Properties Using the Atlas of Australian Soils. Technical Report 11/00. CSIROLand and Water, Canberra ACT.

Milford, H.B., McGaw, A.J.E., Nixon, K.J. (Eds.), 2001. Soil Data Entry Handbook.Third ed. NSW Department of Land and Water Conservation, Sydney.

Mueller, L., Kay, B.D., Hu, Ch., Li, Y., Schindler, U., Behrendt, A., Shephered, G.T., Ball,B.C., 2009. Visual assessment of soil structure. Evaluation of methodologies onsites in Canada, China and Germany. Part I. Comparing visual methods andlinking them with soil physical data and grain yield of cereals. Soil Till. Res. 103,178187.

Nakano, K., Miyazaki, T., 2005. Predicting the saturated hydraulic conductivity ofcompacted subsoils using the non-similar media concept. Soil Till. Res. 84,145153.

Norton, E.A., 1939. Soil Conservation Survey Handbook. USDA, Soil ConservationService, Miscellaneous Publication No. 352. U.S. Government Printing Office,Washington, DC.

Poesen,J., Ingelmo-Sanchez, F., Mucher, H., 1990. The hydrological response of soil

surfaces

to

rainfall

as

affected

by

cover

and

position

of

rock

fragments

in

the

toplayer. Earth Surf. Proc. Landforms 15, 653671.Porta, J., Lopez-Acevedo, M., Roquero, C., 1994. Edafologa para la Agricultura y el

Medio Ambiente. Mundiprensa, Madrid, pp. 645687.Radcliffe, D.E., Rasmussen, T.C., 2002. Physical properties of primary particles.

In: Warrick, A.W. (Ed.), Soil Physics Companion. CRC Press, New York, pp.85126.

Ravina, I., Magier, J., 1984. Hydraulic conductivity and water retention of clay soilscontaining coarse fragments. Soil Sci. Soc. Am. J. 48, 736740.

Rawls, W.J., Brakensiek, D.L., Saxton, K.E., 1982. Estimation of soil water properties.Trans. ASAE 25, 13161320 1328.

Rawls, W.J., Gimenez, D., Grossman, R., 1998. Use of soil texture, bulk density andslope of the water retention curve to predict saturated hydraulic conductivity.Trans. ASAE 41, 983988.

Sauer, Th.J., Logsdon, S.D., 2002.Hydraulic and physicalproperties of stony soils in asmall watershed. Soil Sci. Soc. Am. J. 66, 19471956.

Shafiq, M., Hassan, A., Ahmad, S., 1994. Soil physical properties as influenced byinduced compaction under laboratory and field conditions. Soil Till. Res. 29,1322.

Table 5

Number of predictions against number of observations in each class of Ks according

to the cross-validation.

Predicted/

observed


12/12

Sonneveld, B.G.J.S., Albersen, P.J., 1999. Water erosion assessment on expert knowl-edge and limited information using an ordered logit model. J. Soil WaterConserv. 54, 592599.

The R Development Core Team, 1999. The Comprehensive R Archive Network. .Tietje, O., Hennings, V., 1996. Accuracy of the saturated hydraulic conductivity

prediction by pedo-transfer functions compared to the variability within FAOtextural classes. Geoderma 69, 7184.

U.S. Department of Agriculture (USDA), 1971. Soil Conservation Service. Guide forInterpreting Engineering Uses of Soils. U.S. Government Printing Office,Washington, DC.

U.S. Department of Agriculture (USDA), 2010. National Soil Survey Handbook, Title

430-VI.

Natural

Resources

Conservation

Service

Available

from:

http://soils.usda.gov/technical/handbook (accessed 09.06.10).Verbist, K., Baetens, J., Cornelis, W.M., Gabriels, D., Torres, C., Soto, G., 2009.

Hydraulic conductivity as influenced by stoniness in degraded drylands ofChile. Soil Sci. Soc. Am. J. 73, 471484.

Wahl, N.A., Buczko, U., Bens, O., Huttl, R.F., 2003. Infiltration capacity andmacroporosity of a siltyloamy soil under different tillage systems. In: EGS AGU EUGJoint Assembly, Abstracts from the meeting held in Nice, France, 6-11 April 2003, abstract #5326.

Weiler, M., Naef, F., 2003. An experimental tracer study of the role of macropores ininfiltration in grassland soils. Hydrol. Process. 17, 477493.

Weynants, M., Vereecken,H.,Javaux, M., 2009. Revisiting Vereeckenpedotransferfunctions: introducing a closed form hydraulic model. Vadose Zone J. 8,8695.

Wu, L., Pan, L., Roberson, M.J., Shouse, P.J., 1997. Numerical evaluation of ring-infiltrometers under various soil conditions. Soil Sci. 162, 771777.

Wu, L .,

Pan,

L.,

Mitchell,

J.,

Sanden,

B.,

1999.

Measuring

saturated

hydraulicconductivity using a generalized solution for single-ring infiltrometers. SoilSci. Soc. Am. J. 63, 788792.

Zhang, Sh., Grip, H., Lovdahl, L., 2006. Effect of soil compaction on hydraulicproperties of two loess soils in China. Soil Till. Res. 90, 117125.

http://soils.usda.gov/technical/handbookhttp://soils.usda.gov/technical/handbookhttp://soils.usda.gov/technical/handbookhttp://soils.usda.gov/technical/handbookhttp://soils.usda.gov/technical/handbook

Soil Saturated Hydraulic Conductivity Assessment From Expert Evaluation

Documents

Transcript of Soil Saturated Hydraulic Conductivity Assessment From Expert Evaluation